Adolescent delinquents not only harm their victims, they are at high risk for a host of serious mental and physical health consequences. Because it is very difficult to reduce antisocial behavior after it begins, there is great interest in developing effective methods of preventing delinquency before it starts. Because the onset of antisocial behavior is often in the toddler years, however, early prevention must begin very early in life. Therefore, we need to rigorously evaluate the hypothesized risk factors for delinquency that operate during pregnancy and infancy (e.g., prenatal exposure to substances and parenting during infancy). Crucially, the field must be able to distinguish between non-causal correlates of delinquency and causal risk factors because only causal risk factors are useful targets for prevention. Unfortunately, the field's current understanding of early causal risk factors for delinquency may be seriously flawed for many reasons. Most importantly, the designs used in nearly all previous studies did little to distinguish causal factors from non- causal correlates. Innovation: The field must move beyond traditional between-family designs that compare children from unrelated families and thereby confound hypothesized risk factors with the myriad correlated factors that differ between families. Instead, we propose to adapt the powerful quasi-experimental method of sibling-comparison analysis developed for econometric research to test our hypotheses. This approach compares siblings within nuclear families who vary on the risk factor to determine if they differ in delinquency, greatly minimizing genetic and environmental confounding to provide rigorous tests of causal hypotheses. Approach: We propose to analyze longitudinal data on the offspring of a U.S. national sample of 4,926 women. For each putative risk factor during pregnancy and infancy, we propose to conduct sibling-comparison analyses to test key causal hypotheses: Does each putative causal risk factor influence risk for adolescent delinquency (and do so differently for youth on different developmental trajectories)? Do these early risk variables indirectly influence adolescent delinquency through their influence on childhood conduct problems? Does each putative early causal risk factor influence which children with high levels of childhood conduct problems advance to adolescent delinquency (versus ceasing to be antisocial)? Furthermore, in order to strengthen the empirical basis for understanding the mechanisms of early risk variables, we will determine the extent to which any effects of early risk factors are different for girls and boys and are mediated or moderated by other risk variables in later childhood. Specifically, do early risk factors make some children more vulnerable to inadequate parenting in later childhood and deviant peer influence during adolescence? Alternatively, can adaptive parenting and lack of peer pressure mitigate early risks?